Bioacoustic Extraction System by Modeling Biological Sounds and Optimizing Demixing Vectors

2023 Fiscal Year Final Research Report

• Project/Area Number: 22K18778
• Research Category: Grant-in-Aid for Challenging Research (Exploratory)
• Allocation Type: Multi-year Fund
• Review Section: Medium-sized Section 20:Mechanical dynamics, robotics, and related fields
• Research Institution: The University of Tokyo
• Principal Investigator: Itoh Toshihiro 東京大学, 大学院工学系研究科(工学部), 教授 (80262111)
• Project Period (FY): 2022-06-30 – 2024-03-31
• Keywords: 心音 / 生体音 / 音源抽出 / マイクロフォンアレイ

Outline of Final Research Achievements

The aim of this study is to realize a system to measure heart sounds with minimizing noises such as other biological sounds and ambient noises. Specifically, based on the mechanism of cardiac sound generation, the heart sounds were modeled as a time series signal consisting of four valvular sounds. Demixing vectors that can separate superimposed signals were also introduced. By simultaneously optimizing the parameters of the heart sounds model and the demixing vectors, we successfully developed an algorithm to extract the heart sounds from measured sounds and separate them into the valvular sounds.

Free Research Field

マイクロシステム実装工学

Academic Significance and Societal Importance of the Research Achievements

開発したアルゴリズムは、心音を弁音の状態に分離（復元）でき，これは例えば肺高血圧症の非侵襲診断へ応用できることを示している。また、計測音のチャネル数増加により、環境・生体由来のノイズを全く含まない心音計測へ発展できると考えられ、これは心音データからの心血管疾患自動診断技術の飛躍的な性能向上に繋がる。さらに心音モデル部分を他のモデルに置き換えることで、他の生体音計測を含む包括的なヘルスケアモニタリングや、産業機器の異常音検知など、多方面への発展も可能である。